AIgument

Product Logo
Working Demo
The Team. From left to right; Agam, Bruce, Somya, Teige, Daniel, Clinton

Inspiration

In the age of social media, it feels like we all live in our own echo-chamber of viewpoints. So frequently, we are told how to act and think without ever hearing the other side. But what if it didn't have to be like that? What if you could get a holistic view on any topic, and be able to form your own opinions? And what if it was presented in a far more entertaining way than just reading the news?

What it does

AIgument is a web-based application that generates an argument for and against any topic of your choosing! The conversational tone of the agents and messenger-like UI design reads as if you're watching two friends argue on a Whatsapp chat.

How we built it

The backend was built with Python, FAST API and Athropic's API. The frontend was built with Next.js and ShadCN.

Challenges we ran into

One challenge was only meeting each other on the day the Hackathon began. We had to rapidly build team cohesion and learn to compromise with each other.

Accomplishments that we're proud of

We are very proud to have built such a clean product, that seamlessly connects the frontend and backend. Not only that, but we are proud of our pitch and confident in our vision for this product.

What we learned

We learnt many things from each-others previous coding experiences, including full-stack technologies, high-value prompt engineering skills, working with people with different skill sets, and delivering a project in a short time frame.

What's next for AIgument

There are many features we wish to add to AIgument, including (but not limited to):

Sign In and User Authentication
Saved Chats
Giving the AI agents access to recent news
Allow users to tailor the AIs' tones and conversational style (informal -> formal, casual -> structured debate, calm -> angry)
Podcast mode (using Text-to-Speech models)
A mode for users to debate with the AI (with feedback on the user's arguments)

Built With

claude
fastapi
next.js
python
shadcn
typescript

Submitted to

UNIHACK 2026, funded by the European Union

Created by

I was the product leader and head of marketing. As the group went with my idea, I led the analysis and design phases in Day 1, before helping out with the code where necessary. On Day 2, I focussed on guiding the group towards an MVP and the pitch video.

Teige Cordiner
I worked on Front End using NextJS, built a fundamental page structure and finished deployments & integrations

Bruce Huang
I worked on connecting the backend to the frontend, designing the endpoints and deploying the backend on vercel.

Agam Singh
I worked on the design. (and alittle front-end) This was my first time using pnpm and React. I have had exposure to React during a typescript/haskel class, however that was not changing any of the supplied React. Also, I learned that CSS Libraries / React Libraries exist

Clinton Sibberas
As the Backend Reliability Engineer, I focused on the 'Zero-Hallucination' architecture of the project. While the team developed the initial API and frontend integration, I engineered a Multi-Agent Verification Loop to ensure the integrity of the AI's arguments.

My specific contributions included building the automated fact-checking agent and implementing a self-correction retry loop in FastAPI

Sam Mittal
I specialised in designing high-performance prompts for Claude AI agents using the Anthropic API and Python, optimising for response quality. On the architecture side, I designed and implemented a FastAPI layer bridging frontend and backend systems, and led significant backend development — including AI configuration and endpoint design for seamless integration.

Daniel Pavela